edelwace-0.1.0.0: HaskTorch Reinforcement Learning Agents for GACE

Safe Haskell	None
Language	Haskell2010

RPB.ERE

Description

Emphasizing Recent Experiences

Synopsis

samplingRange :: Int -> Int -> Int -> Int -> Float -> Int
sample :: Buffer Tensor -> Int -> Int -> Int -> Int -> Int -> Float -> IO (Buffer Tensor)
anneal :: Float -> Float -> Int -> Int -> Float

Documentation

samplingRange Source #

Arguments

:: Int	Buffer Size N
-> Int	Number of Epochs K
-> Int	Current Epoch k
-> Int	cMin
-> Float	η
-> Int	cK

Calculate ERE Sampling range cK

sample :: Buffer Tensor -> Int -> Int -> Int -> Int -> Int -> Float -> IO (Buffer Tensor) Source #

Sample for buffer within ERE range

anneal Source #

Arguments

:: Float	Initial η0
-> Float	Final ηt
-> Int	Horizon T
-> Int	Current step t
-> Float	Current ηt

ERE η Annealing during training